OpenCode Go in Codex — Use the Go Subscription with Codex Router

OpenCode Go is a flat-rate subscription that gives you a generous monthly allowance of top open-source coding models — Grok, GLM, Kimi, DeepSeek, Qwen, MiniMax, MiMo and more — for $10/month ($5 for the first month). It is designed to be used with any agent, and Codex Router is the official way to use it inside the Codex App and CLI: one API key, and the Go models appear in your normal model picker.

OpenCode Go$10/monthflat-rate

Why OpenCode Go is a good deal for Codex users

Go charges one flat fee instead of per-token API costs. The allowance refreshes every five hours, and even the premium models have generous room:

The exact allowance depends on the tier you pick (1×, 10×, 25×, 50× and higher). For heavy Codex use, a subscription usually costs less than paying per token for the same models through their vendor APIs — and it removes the per-request pricing anxiety entirely.

Go vs Zen

OpenCode has two billing models, and they are not interchangeable:

Codex Router supports both: the opencode-go provider family covers the Go subscription, and opencode-zen covers the pay-per-use endpoint.

Use Go in Codex with Codex Router

Install Codex Router first (see the install guide), then set the Go API key and enable the provider family:

./bin/model-router codex provider-key opencode-go set
./bin/model-router codex providers enable opencode-go

The Go API key comes from your OpenCode account at opencode.ai/auth — the same key the official docs use for the /connect flow, and the same one Codex Router reads from its protected key file. There is no separate “Codex” key; one OpenCode account key covers both.

On Windows use ./model-router.ps1 codex with the same commands. Then fully quit and reopen Codex — the Go models appear in the picker next to your native GPT models, with the usual sub-agent, Vision Bridge and credential-isolation behavior.

The key is stored in a protected per-user file (mode 600 on POSIX, current-user ACL on Windows) and is only ever sent to opencode’s own endpoints.

Models included in the Go subscription

These 19 registry-proven models ship with the router and appear in the picker once the provider is enabled:

opencode-go carries the Chat Completions models, opencode-go-messages the Anthropic Messages models and opencode-go-responses the Responses models — all four are one selectable family sharing a single stored key. Set OPENCODE_GO_BASE_URL only if you need to point at a different endpoint.

Fair use and limits

• The allowance refreshes every 5 hours, not per calendar day.
• “2× usage” models (DeepSeek V4 Flash, GPT 5.6 Luna) draw from the doubled tier when you select it.
• You can top up credit if you run out, and cancel any time.
• Codex traffic through the router consumes the same Go allowance as any other agent.

Get OpenCode Go

OpenCode Go costs $10/month with a $5 first month. If you subscribe through our referral link, both you and we receive $5 of Go usage credit.

FAQ

Do I need an OpenCode account to use Go in Codex? Yes — the Go subscription is tied to an OpenCode account. Create one, subscribe to Go, and the same account’s API key works in Codex Router.

Does using Go through Codex Router count against the same allowance? Yes. OpenCode bills Go by requests to its endpoints, and the router simply relays your Codex requests there — same account, same 5-hour allowance.

Can I use Go and my ChatGPT login at the same time? Yes. Native GPT models stay on your ChatGPT sign-in, and Go models route through the subscription. The router keeps both credentials isolated.

What if I already have vendor API keys? Entries like DeepSeek V4 Pro intentionally coexist — a deepseek/deepseek-v4-pro key is billed by DeepSeek, while opencode-go/deepseek-v4-pro is billed by your Go subscription. Pick whichever billing you prefer.

Related

• Supported models — the full provider and model matrix
• Installation guide — set up Codex Router first
• How it works — catalog merging, routing and credential isolation